The study was carried out in the period January-May 2016 inside the Botanical Garden of Rome (41˚53'53'' N, 12˚28'46'' E; 53 m a.s.l.). The Botanical Garden covers an area of 12 ha in the city centre, between Lungara Street and the Gianicolo Hill. The plane area is enriched with tree species, the Palm Collection and meadows, while the hill area is occupied by Ferns, Eucalyptus collection, Bamboos, Rose Garden, Japanese Garden, Rock Garden and Geophytes, Mediterranean Wood and Gymnosperms [21] .

The study area is under a Mediterranean type of climate. The average total annual rainfall is 848 mm, most of it distributed in autumn and winter. The average maximum air temperature of the hottest months (July and August) is 31.7˚C ± 0.1˚C and the average minimum air temperature of the coldest month (January) is 4.9˚C ± 0.9˚C. The mean yearly air temperature is 16.7˚C ± 6.5˚C (data provided by the Lazio Regional Agency for Development and Agricultural Innovation; Meteorological Station of Rome, Lanciani Street, data for the period 2006 to 2016).

Atmospheric carbon dioxide concentration (CO₂, ppm), air temperature (T_a, ˚C) and air humidity (RH, %) were monitored simultaneously by handheld tools (Rotronic, CP11) along two Transects: Transect 1 from Garibaldi Square to the centre of the Botanical Garden and Transect 2 from Lungotevere della Farnesina Street to the centre of the Botanical Garden. Measurements were carried out in three different points of each transect: at the centre of the Garden (C), at 150 m from the centre (B₁ and B₂, for Transect 1 and 2, respectively) and outside the Botanical Garden (at 300 m from the centre, A₁ at Garibaldi Square and A₂ at Lungotevere della Farnesina, for Transect 1 and 2, respectively) (Figure 1). Traffic level (i.e. number of vehicles per minute) was monitored simultaneously with CO₂ measurements in Garibaldi Square and Lungotevere della Farnesina Street. The sites along each transect were chosen on the base of a progressive traffic intensity decrease from street densely congested to the inner of the Botanical Garden.

The most important plant collections developing in the Botanical Garden were considered. In particular, the Mediterranean Garden (2050 m²), Palms (4463 m²), Gymnosperms (15,500 m²), Bamboos (6205 m²), Mediterranean Wood (17,850 m²), Eucalyptus Collection (8500 m²), Japanese Garden (2250 m²), Erythrina Area (13,005 m²), Ferns (4250 m²), Rock Garden and Geophytes (5100 m²), Rose Garden (4250 m²) and Meadows (2423 m²) (Figure 2).

The extension of each plant collection was measured by a Quantum Gis (QGIS), an Open Source Geographic Information System (OSGEO4W, version 1.8.0) running on Windows. QGIS determines the acquisition, recording, analysis, visualization and restitution of information by geographical data. The GIS

software is useful for the census of urban green areas by the analysis of digital cartographies.

Leaf Area Index (LAI) was measured by the “LAI 2000 Plant Canopy Analyzer” (LICOR Inc., Lincoln, USA) for all the considered plant collections. Structural traits of each collection, excluding Meadows, were measured on representative plants (n = 10 per plant collection). In particular, tree diameter at breast height (DBH, m) was measured by callipers (Silvanus calliper—65 cm), and a DBH tape (length = 20 m) when diameter was larger than 65 cm. Plant height (H, m) was measured by electronic clinometers (Haglöf, Sweden). The total photosynthetic leaf surface area (TPS, m²) of each plant collection was determined by multiplying each LAI value by the extension of the plant collection.

The CO₂ sequestration capability for each plant collection was calculated multiplying TPS by the mean yearly net photosynthesis and the total yearly photosynthetic activity time (in hours), according to [22] . In order to compare CO₂ sequestration capability of the different plant collections, the CO₂ sequestration capacity per hectare (CS, Mg CO₂ ha⁻¹∙year⁻¹) was calculated. The total CO₂ sequestration capacity of the Botanical Garden was also calculated (CS_Tot).

The net photosynthetic rate (N_P, µmol∙m⁻²∙s⁻¹) was measured by an open infrared CO₂ gas analyzer (ADC LCPro+, UK), equipped with a leaf chamber (PLC, Parkinson Leaf Chamber). Measurements were made in situ on cloud-free days (PAR > 1000 µmol∙m⁻²∙s⁻¹), in the morning (from 9:00 am to 12:00 pm), to ensure that near maximum daily N_P was measured. On each sampling occasion, fully sun expanded leaves were used (Varone et al., 2015). Leaves were retained in their natural position during measurements. Measurements were carried out at ambient air temperature on five representative plants per each collection (three leaves per plants).

The monetary value of CO₂ sequestration for the collections growing in the Botanical Garden was estimated assuming a monetary value of $ 0.00334/lb (i.e. $0.00736/kg) for sequestered CO₂, according to [23] .

Differences of the means were tested by one-way analysis of variance (ANOVA), and Tukey test for multiple comparisons. Moreover, in order to understand how LAI, TPS and N_P (predictors) affected CS (response variable), the predictors were combined via Principal Component Analysis (PCA) across plant collections. Then, a simple linear regression analysis between the axis explaining the largest proportion of the variance (i.e. PC₁) and CS was carried out. All statistical tests were performed using a statistical software package (Statistica, Statsoft, USA).

The mean CO₂ concentration during the study period was 485 ± 22 ppm (mean value of Transect 1 and Transect 2) peaking in January (522 ± 21 ppm, mean of A₁ and A₂) (Table 1). The highest CO₂ concentration was monitored where the traffic level was the highest (31 ± 1 vehicles∙min⁻¹, mean of A₁ and A₂) decreasing by 8% in May (mean of A₁ and A₂) associated to a 21% decrease of the traffic level (mean of A₁ and A₂). Along the Transect 1, A₁ had a 6% lower CO₂ concentration compared to A₂ (Transect 2) with a 70% lower traffic level. Moreover, CO₂ concentration decreased, on average, 9% from outside to the centre of the Botanical Garden in both the Transects. In particular, CO₂ concentration decreased 2% and 6% from A₁ to B₁ and from A₁ to C, respectively, along Transect 1, and 6% and 12% from A₂ to B₂ and from A₂ to C, respectively, along Transect 2.

A similar trend was observed in T_a decreasing, on average, 14% from the outside to the centre of the Botanical Garden. In particular, T_a decreased 7% and 13% from A₁ to B₁ and from A₁ to C, respectively, along Transect 1, and 8% and 14% from A₂ to B₂ and from A₂ to C, respectively, along Transect 2. An opposite trend was observed for RH, increasing, on average, 7% from outside to inside the Botanical Garden (Table 1).

Structural traits of plant collections are shown in Figures 3(a)-(d). The highest LAI (p ≤ 0.05) was measured in Bamboos (2.90 ± 0.06) and the lowest in Meadows (0.40 ± 0.05, mean value). The Mediterranean Wood had the highest (p ≤ 0.05) TPS (43,197 m²) and the Meadows the lowest (969 m²). DBH was the highest (p ≤ 0.05) in the Gymnosperms (1.16 ± 0.30 m) and the lowest in Rose Garden (0.06 ± 0.03 m). The Eucalyptus Collection had the highest (p ≤ 0.05) H (26.8 ± 2.3 m).

Bamboos had the highest CS (210 Mg CO₂ ha⁻¹∙year⁻¹, corresponding to 57.4 Mg C ha⁻¹∙year⁻¹) among plant collections, followed by the Mediterranean Wood (133 Mg CO₂ ha⁻¹∙year⁻¹, corresponding to 36 Mg C ha⁻¹∙year⁻¹) while Ferns had the lowest CS (7.1 Mg CO₂ ha⁻¹∙year⁻¹, corresponding to 1.9 Mg C ha⁻¹∙year⁻¹) (Figure 4). Considering the extension of each plant collection, the Mediterranean Wood had the highest CO₂ sequestration capability (150 Mg CO₂ year⁻¹, corresponding to 65 Mg C year⁻¹), followed by the Bamboos (130 Mg CO₂ year⁻¹, corresponding to 36 Mg C year⁻¹), the Erythrina Area (110 Mg CO₂ year⁻¹, corresponding to 30 Mg CO₂ year⁻¹), Gymnosperms (109 Mg CO₂ year⁻¹,

corresponding to 29 Mg CO₂ year⁻¹), Eucalyptus Collection (35 Mg CO₂ year⁻¹, corresponding to 9 Mg CO₂ year⁻¹), Palms (28 Mg CO₂ year⁻¹, corresponding to 8 Mg CO₂ year⁻¹), Mediterranean Garden (17 Mg CO₂ year⁻¹, corresponding to 5 Mg CO₂ year⁻¹), Japanese Garden (15 Mg CO₂ year⁻¹, corresponding to 4 Mg CO₂ year⁻¹), Rose Garden, Rock Garden and Geophythes (10.4 ± 0.5 Mg CO₂ year⁻¹, mean value, corresponding to 2.8 ± 0.1 Mg CO₂ year⁻¹), Meadows (4.7 Mg CO₂ year⁻¹ corresponding to 1.3 Mg C year⁻¹) and Ferns (3.0 Mg CO₂ year⁻¹, corresponding to 0.82 Mg C year⁻¹). Considering the total extension of plant collections (8.6 ha), the CO₂ sequestration capability for the Botanical Garden was of 6947 Mg CO₂ year⁻¹, corresponding to 1897 Mg C year⁻¹.

The PCA returned two axis of variations across plant collections. In particular, PC₁ was positively related to LAI and TPS accounting for 55% of the total variance. PC₂ was positively related to N_P accounting for 37% of the total variance. There was a significant linear regression between PC₁ and CS. Nevertheless, the relationship did not hold for Gymnosperms and Bamboos collections which felt apart from the fitted line. When they were removed from the analysis, the performance of the linear model significantly increased (Figure 5).

The monetary value of CO₂ sequestered by the Botanical Garden was 59-56$ ha⁻¹∙year⁻¹, to which Bamboos and Mediterranean Wood gave the highest contribution (26% and 16%, respectively), and Ferns, Rose Garden (1%) and Meadows (2%) the lowest.